Rede neural para correção de sinais distorcidos pela saturação de transformadores de corrente

Abstract

This work presents an algorithm that aims to reduce the effects of saturation of current transformers (CTs), being one of the factors that affect the performance of protection systems in the electric power systems. For this end, the paper proposes the use of a methodology to correct the distorted portions of secondary current signals, provided by saturated CTs, based on the use of Artificial Neural Networks (ANN). This technique uses only one neuron to determine the characteristic parameters of the short-circuit current signal, and then apply them to correct the distorted ranges present in the referred signal waveform. Cases of short-circuits on a transmission line were simulated, where the CTs operate on different scenarios previously defined, for which, depending on the levels of fault currents, saturation occurs in its core. These simulations were performed to generate signals to be processed by the proposed method, considering different fault conditions. These simulations were performed using the software EMTP-ATP. Saturated signals were corrected by the algorithm developed, which was implemented in MATLAB ®, using proper function. From the generated signals processing by the proposed technique, one can verify the good performance of the proposed methodology, considering the correction accuracy and the ease of implementation of the referred methodology. With this proposal, it is expected to provide an alternative to protection schemes against the situations involving the saturation of current transformers, ensuring selective and coordinated protective relaying, increasing the reliability of the electric power systems.